# HPs optimizer_choice = int(sys.argv[batch_size_index + 1]) arg1 = float(sys.argv[batch_size_index + 2]) # lr arg2 = float(sys.argv[batch_size_index + 3]) # momentum arg3 = float(sys.argv[batch_size_index + 4]) # weight decay arg4 = float(sys.argv[batch_size_index + 5]) # dampening dropout_rate = float(sys.argv[batch_size_index + 6]) activation = int(sys.argv[batch_size_index + 7]) # Load the data print('> Preparing the data..') if dataset is not 'CUSTOM': dataloader = DataHandler(dataset, batch_size) image_size, number_classes = dataloader.get_info_data trainloader, validloader, testloader = dataloader.get_loaders() else: # Add here the adequate information image_size = None number_classes = None trainloader = None validloader = None testloader = None # Test if the correct information is passed - especially in the case of CUSTOM dataset assert isinstance(trainloader, torch.utils.data.dataloader.DataLoader ), 'Trainloader given is not of class DataLoader' assert isinstance(validloader, torch.utils.data.dataloader.DataLoader ), 'Validloader given is not of class DataLoader' assert isinstance(testloader, torch.utils.data.dataloader.DataLoader ), 'Testloader given is not of class DataLoader'